Measured Zout
We measured Zout today. It came out a little higher than I expected at 1.67 Ohms.
By my calculations, copper losses in the OT account for ~0.55 Ohms (secondary plus primary reflected to secondary). So it looks like the output tube's effective plate resistance is ~650 Ohms or so.
We measured by taking a signal generator and injecting a signal through a 50 Ohm resistor into the amplifier output, then solved for the unknown resistance. It was fun to reason it out with the kids and watch them solve the problem.
We measured Zout today. It came out a little higher than I expected at 1.67 Ohms.
By my calculations, copper losses in the OT account for ~0.55 Ohms (secondary plus primary reflected to secondary). So it looks like the output tube's effective plate resistance is ~650 Ohms or so.
We measured by taking a signal generator and injecting a signal through a 50 Ohm resistor into the amplifier output, then solved for the unknown resistance. It was fun to reason it out with the kids and watch them solve the problem.
Are you proposing applying feedback from the OT secondary or did you have something else in mind?
Looking at the original schematic in post #1, I'm seeing two N Feedback loops.
The obvious one is the so called "Shunt Schade" from output plate to output grid driven by current drive.
The second hidden N Fdbk loop does involve the output tube and the Op Amp thru resistor R1. Current derived from the output plate voltage and R2 passes thru the Mosfet down thru R1. (ie, same plate V derived current thru R2 and R1) The voltage on R1 reflects an attenuated plate output voltage, and is being fed back to the Op Amp inverting input, providing an outer loop N Fdbk. This is really quite ingenious. The Mosfet avoids any NPN base current distortion effect besides.
The presence of the hidden outer driver loop N Feedback is surprising. One starts to wonder what the RCA 50 Watt Amp has going on with an extra driver CFB loop in addition.
This can be generalized to an all tube solution easily. Just put an LTP of pentodes in place of the Op Amp and the Mosfet, with a resistive tail or CCS tail. (optional floating screen supplies from tail at this point). The LTP output pentode just gets a constant bias V on it's grid. The LTP output N Fdbk current will have to be complementary to the input LTP pentode signal current.
If the screen supplies are not cathode floated, but just summed screen grid currents to ground (single screen V supply from ground), then the LTP will have gain droop at signal extremes when plate V drops below screen V. This might be useful for soft limiting. Have to experiment here.
Then one can go to a full P-P scheme by just filling out the symmetry. Both LTP driver pentodes are used as complementary "shunt Schade" drivers to P-P output tubes, with complementary inverted inputs into the two LTP pentode grids. Anatoliy should like the hidden nested loops here. An outer global loop would be 3 loops.
Note: The RCA 50 Watt design (below) uses a pull-up resistor (R15, R19) on the driver plates to B+. This inserts an additional current component in the driver plate output. The driver CFB loops would provide an automatic tracking increase of driver current to compensate, if sized correctly. Maybe this is what they are doing there. Would have to calculate out the effects to see if they are equal. Very impressive design if that's what it's doing, no wonder nobody could figure out what they were up to.
The obvious one is the so called "Shunt Schade" from output plate to output grid driven by current drive.
The second hidden N Fdbk loop does involve the output tube and the Op Amp thru resistor R1. Current derived from the output plate voltage and R2 passes thru the Mosfet down thru R1. (ie, same plate V derived current thru R2 and R1) The voltage on R1 reflects an attenuated plate output voltage, and is being fed back to the Op Amp inverting input, providing an outer loop N Fdbk. This is really quite ingenious. The Mosfet avoids any NPN base current distortion effect besides.
The presence of the hidden outer driver loop N Feedback is surprising. One starts to wonder what the RCA 50 Watt Amp has going on with an extra driver CFB loop in addition.
This can be generalized to an all tube solution easily. Just put an LTP of pentodes in place of the Op Amp and the Mosfet, with a resistive tail or CCS tail. (optional floating screen supplies from tail at this point). The LTP output pentode just gets a constant bias V on it's grid. The LTP output N Fdbk current will have to be complementary to the input LTP pentode signal current.
If the screen supplies are not cathode floated, but just summed screen grid currents to ground (single screen V supply from ground), then the LTP will have gain droop at signal extremes when plate V drops below screen V. This might be useful for soft limiting. Have to experiment here.
Then one can go to a full P-P scheme by just filling out the symmetry. Both LTP driver pentodes are used as complementary "shunt Schade" drivers to P-P output tubes, with complementary inverted inputs into the two LTP pentode grids. Anatoliy should like the hidden nested loops here. An outer global loop would be 3 loops.
Note: The RCA 50 Watt design (below) uses a pull-up resistor (R15, R19) on the driver plates to B+. This inserts an additional current component in the driver plate output. The driver CFB loops would provide an automatic tracking increase of driver current to compensate, if sized correctly. Maybe this is what they are doing there. Would have to calculate out the effects to see if they are equal. Very impressive design if that's what it's doing, no wonder nobody could figure out what they were up to.
Attachments
Last edited:
Next phase will be to build a negative rail for tube solution comparisons.
First step will be a simple pentode driver (probably 6AU6) but differential stage will follow. I have 6BN11s waiting to jump in there. Maybe I'll even recreate an RH-style triode driver just so I can show what it does.
I have various ideas and the goal will be to finally answer in my mind which are better and by how much. I'm trying to follow something like the Douglas Self method of amp design and only add complexity where it makes a significant difference to system-level performance, so I have to compare some options in a real amp.
A negative rail for the LTP driver tail would avoid the need for the driver plate pull-up resistors as RCA used (R15, R19). So no need for the EXTRA driver CFB pathes complexity (well, assuming a driver load R current nulling solution was what they did). The "shunt Schade" resistors can provide sufficient voltage to the driver stage then using a neg. rail. An Ultra-efficient parts design, providing 3 nested N FDbks with just 2 resistors (if global N Fdbk is included too).
Last edited:
It gives a good experience to design by sound. As the bass is not good , you can show how by overall feedback ( you need 10db ) the sound enhances in bass yes, but the voices loose realisme. You can also show positive current feedback how it reduces the output impedance, by applying an adjust, you can show how passing from negative to positive current feedback the sound alters. Once the problem sorted out, you can review the design , by higher operating current for example . Keep the children all along the design in mind, that the amplifier is to satisfy the ears , not the measuring equipment.
Last edited:
the G2 supply consisted of an IXYS01N100D CCS set to around 30ma and shunted by two 75v 1W zeners, this is believed to be the reason why the amp sounded so good.....G2 supply is 150 for both the tubes...
Last edited:
Tony, is that 150V G2 for the outputs and ground referenced for the drivers too, or is it a floating supply on the driver cathode tail? (like Pete's circuit below with the 0A2 regulator)
Push-pull driver board
Looks like Pete's driver board could be easily modded to do this design, with no driver pull-up resistors (remove R35, R4). The neg. rail V needs to be sufficient to operate the pentode drivers thru the shunt Fdbk resistors (R1, R39), and the pentode driver grids get re-referenced to near the neg. V.
One might like to put a signal splitter in ahead of the pentode drivers for better symmetry.
Last edited:
I swapped it out with another and that one is working great.
We did have a brief "red plate incident" during the first attempt at adjusting bias. I was hoping no permanent damage but maybe not...
i had the same happened to me using the 6CL6 in a pentode voltage amp stage and 100k grid leak, i did not investigate further as i just replaced it with a 6bq5 tube instead...
if i may add, the B+ take off point was at the junction of the filter caps in the full wave voltage doubler so that the the raw b+ of 180 volts is then dropped to 150 volts, so the regulator is not dissipating a lot...
Your bottle neck of shunt feedback is the grid 100k , reducing the ratio to 30%. You can use simulated inductor , gyrator as called here. see example here. Extra power,higher DF,lower distortion from SE pentodes
I guess I had convinced myself it would have less of an impact, but it is pretty clear from the measurements I have taken that this is operating at about 30%. Zout would be lower if I were operating at a higher feedback ratio than that.
My longer-term plan for this design approach is to incorporate a smaller transmitting triode as the output tube. I want to find a way to take one of the high-impedance types that aren't otherwise terribly useful in audio and make them perform very well. Tubes like the 811A could make an amp that is quite a bit better than the average 300B amp with this approach.
This will necessitate a grid driver inside the loop to drive the grid positive. I will use a mosfet, and the gate bias resistor can be quite a bit higher than 100k for that. So I'll probably implement the grid driver now on the 6384, then I'll experiment with different 1st stage circuits, and lastly I'll swap in the triode output tube.
I took distortion measurements last night. I'll get them over to this computer soon to post here.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.